Personality and Trust Fosters Service Quality

Personality Predictors of Extreme Response Style

Bobby D. Naemi,1 Daniel J. Beal,1 and Stephanie C. Payne2

1Rice University2Texas A & M University

ABSTRACT Extreme response style (ERS) refers to the tendency tooveruse the endpoints of Likert-type scales. This study examined the extentto which ERS is accounted for by measures of personality, specifically,intolerance of ambiguity, simplistic thinking, and decisiveness. One hun-dred and sixteen pairs of undergraduate students and one of their respec-tive peers completed a battery of questionnaires assessing these personalitymeasures, alongside three measures of extreme responding. Results indi-cate that peer ratings of intolerance of ambiguity and simplistic thinkinginteract with the primary participant’s time spent on the survey to predictthe primary participant’s extreme responding. Thus, those who quicklycomplete surveys and are intolerant of ambiguity or are simplistic thinkersare most likely to exhibit ERS. These results have implications not only forsurveys using rating scales, but also illustrate how epistemic personalityfactors more generally influence the processing of new information.

Extreme response style (ERS) refers to the tendency to dispropor-tionately favor the endpoints or extreme categories of ordinal re-

sponse or Likert-type scales, irrespective of particular item content.ERS can be described as the opposite of central tendency and differs

from other response styles such as acquiescence (yea-saying) andsocial desirability in that most research has found this style remains

relatively consistent across differing construct measures over time(Greenleaf, 1992b; Hamilton, 1968; Jain & Agrawal, 1977; Merrens,1970). Extreme responders represent between 25% and 30% of all

respondents in most surveys (Austin, Deary, & Egan, 2006; Eid &

A previous version of this paper was presented at the 2007 meeting of the Society

for Industrial and Organizational Psychology. The authors thank Faith Dorsey and

Kendra Ross for their help with data collection.

Correspondence concerning this article should be addressed to Bobby Naemi,

Department of Psychology, Rice University, 6100 Main Street, Houston, TX 77005.

E-mail: [email protected].

Journal of Personality 77:1, February 2009r 2009, Copyright the AuthorsJournal compilation r 2009, Wiley Periodicals, Inc.DOI: 10.1111/j.1467-6494.2008.00545.x

I:/Bwus/jopy/545/[email protected]

Rauber, 2000). Failing to account for differences in ERS results in

confounded or spurious correlations, highlighting the practical im-portance of studying extreme responding in personality and survey

research (Greenleaf, 1992a; Johnson, 2003).As most research has shown that ERS occurs consistently over

time, it seems likely that there are certain stable personality traitsthat are characteristic of extreme responders. In short, ERS might be

considered a behavioral manifestation of stable personality traitsthat differ among individuals. As such, it is beneficial to determine

the dispositional and contextual antecedents of ERS in order tobetter predict who is most likely to employ this confounding patternand when it is most likely to be used. Given that most of the research

indicates that ERS is stable, researchers have attempted to identifydispositional antecedents of ERS by exploring the relations between

ERS and various individual difference variables (e.g., Austin et al.,2006; Lewis & Taylor, 1955).

Individual Differences and Extreme Response Style

ERS has been related to various demographic, cultural, and person-

ality variables, but very few studies have offered any sort of theo-retical explanation for why ERS might occur. Some studies haveshown that ERS can differ by sex (Berg & Collier, 1953; Borgatta &

Glass, 1961; Crandall, 1973; Eid & Rauber, 2000) with women en-gaging in ERS more than men, although several studies have failed

to find any sex differences (Brengelmann, 1960b; Greenleaf, 1992b;Light, Zax, & Gardiner, 1965). Additional studies, but not all (e.g.,

Greenleaf, 1992b), have found differences in ethnicity (Bachman &O’Malley, 1984; Gerardo, Gamba, & Marin, 1992).

Some individual differences research also suggests a potentialcognitive ability component to ERS (i.e., lower cognitive ability in-

dividuals engage in more ERS); however, studies investigating rela-tionships between ERS and cognitive ability have produced mixedconclusions, with some research finding negative relations (Breng-

elmann, 1960a; Das & Dutta, 1969; Light et al., 1965; Wilkinson,1970), and others finding none (Kerrick, 1954; Zuckerman & Nor-

ton, 1961). In terms of personality, ERS has been found to be relatedto anxiety (Lewis & Taylor, 1955), extraversion, and agreeableness

(Austin et al., 2006), although no theoretical reasons were providedfor these relations.

2 Naemi, Beal, & Payne

To summarize, most research relating ERS to individual differ-

ence variables seems to be mixed and inconclusive, a result that maybe attributed to Greenleaf’s (1992b) contention that past research is

plagued by inconsistencies in the measurement of ERS. In addition,previous studies examining personality predictors may have been

equivocal due to the lack of a theoretical basis for their predictions.Indeed, in a review of the past literature, Hamilton (1968) found that

ERS did not exhibit any consistent patterns to standard global per-sonality inventories (e.g., 16-PF). We suggest that such broad-stroke

approaches to understanding the personality basis of ERS may notmeet with success. Instead, understanding the nature of extreme re-sponses and motivations people have for making them should nar-

row the field of potential predictors. Specifically, we believe that themost promising dispositional predictors of ERS belong to a cluster

of so-called epistemic variables (Kruglanski, 1989). These personal-ity constructs detail the manner in which people process and respond

to information. Of the several traits discussed in lay epistemic the-ory, perhaps the most notable in its implications for ERS is the

construct of intolerance of ambiguity.

Intolerance of Ambiguity and Extreme Response Style

Intolerance of ambiguity was formally defined by Budner (1962) as

‘‘the tendency to perceive ambiguous situations as sources of threat’’(p. 29). Budner (1962) also defined an ambiguous situation as ‘‘one

which cannot be adequately structured or categorized by the indi-vidual because of the lack of sufficient cues’’ (p. 30). Intolerance of

ambiguity represents a compelling potential predictor of extreme re-sponding, because individuals with a high intolerance of ambiguity

should avoid making equivocal responses and instead gravitate to-ward definite unambiguous choices. As such, high intolerance for

ambiguity could explain why a respondent would avoid using po-tentially ambiguous middle categories on a rating scale and insteadmake disproportionate use of the definite unambiguous endpoints of

a rating scale. Given the research indicating that ERS seems to be astable trait, varying between individuals, it seems likely that a stable

personality construct such as intolerance of ambiguity may act as anexplanatory individual difference variable for extreme responding.

Although intolerance of ambiguity has been suggested as a cor-relate of ERS (Soueif, 1958), there are no studies that have directly

Extreme Response Style 3

related ERS and intolerance of ambiguity using measures uncon-

taminated by other related constructs. For example, Brengelmann(1960a) used only positive extreme responses as a measure of ERS,

and Brim and Hoff (1957) used a measure of need for certainty as aproxy of intolerance of ambiguity. As such, the measure of intoler-

ance of ambiguity used in this study will be designed to tap onlyintolerance of ambiguity and not similar constructs such as social

dominance, rigidity, or need for certainty.

Simplistic Thinking and Extreme Response Style

Another epistemic construct that could potentially be related to ERS

is the tendency to think simplistically or in dichotomized terms. Al-though intolerance of ambiguity reflects a preference for simplicity

and discomfort with complexity, the construct of simplistic thinkingrefers to a tendency to view the world in simplified or dichotomized

terms. Prior research has demonstrated that separating out the ten-dency and preference components of epistemic variables is a useful

distinction and may carry some independent predictive validity (e.g.,Roets & Van Hiel, 2007). The described tendency for simplistic

thinking is reflected in such items as ‘‘I tend to see most issues inblack and white terms’’ and ‘‘Most questions can be answered with asimple yes or no.’’ Responses to items such as these theoretically

should be related to ERS, given that ERS can essentially be definedas the dichotimization of a Likert scale into a simple yes/no re-

sponse. In this sense, a tendency to think in simple or reductive termswould drive a respondent to employ primarily the endpoints of a

scale as a way of simplifying potential responses. As such, a dispo-sitional tendency for simplistic thinking is also expected to predict

extreme responding.

Decisiveness and Extreme Response Style

A third personality predictor that may provide a theoretical mech-

anism underlying ERS is the construct of decisiveness. Decisivenesshas been globally defined as encompassing the ease, speed, and con-

fidence of making firm decisions (Kruglanski, 1989; Thompson,Naccarato, & Parker, 1989). A dispositional tendency to quickly

seize on a solid answer that provides the strongest possible decisionmight lead respondents to commit to a definite extreme response,


whereas indecisiveness may lead to respondents avoiding extreme

categories that represent strong decisive opinions. This hypothesizedrelation has some precedence in cross-cultural research relating mas-

culinity and ERS. Johnson et al. (2005) posited that features ofmasculine cultures that emphasize assertiveness and decisiveness

might lead respondents to choose the strongest available categoriesto express their opinions, and the researchers subsequently found a

significant relation between masculinity and extreme responding.1

Therefore, we also expected decisiveness to predict individual differ-

ences in ERS.Each of these three personality constructs, intolerance of ambi-

guity, simplistic thinking, and decisiveness therefore should be pos-

itively related to extreme responding. Although these three measuresbelong to the overarching category of epistemic constructs, they do

so for varying theoretical reasons. As such, we expected all three toaccount for unique variance in ERS.

Contextual Effects on Extreme Response Style

Although we assert that ERS is a behavior linked primarily to stable

personality traits, it is unlikely that people are completely uniform intheir tendency to engage in ERS. Much like any trait, people will

undoubtedly vary in their expression of the trait from situation tosituation (Mischel & Shoda, 1995). Therefore, we examined two po-

tential contextual effects on ERS. Hui and Triandis (1985) proposedthat ERS was more likely to occur at the end of a questionnaire due

to fatigue or boredom with the measure, highlighting one potentialcontextual effect. We accounted for this possibility by varying thelocation of certain measures of ERS throughout the questionnaire

battery. Examining extreme responses as a function of the order inwhich the scales are presented provides a check for the fatigue hy-

pothesis: If extreme responses to the same content are more likely atthe end of a questionnaire, then fatigue is a likely explanation. If

there are no differences, however, then some evidence is provided forthe stability of ERS.

Another potential contextual influence on ERS involves the con-servation of cognitive resources when completing surveys. Specifi-

cally, people who spend less time on a survey might be more likely to

1. Interestingly this finding runs counter to past research suggesting women are

more likely to make extreme responses than men.


make extreme responses, simply because extreme responses are the

simplest and least time-consuming choice to make when differenti-ating among Likert scale categories, given that an extreme response

is essentially the equivalent of Yes or No, or Agree or Disagree. Inthis sense, time taken to complete a survey, which has also been

found to be related to trait impulsivity (Molto, Segarra, & Avila,1993), may be related to extreme responding.

It is also possible, however, that people could rush through a sur-vey because they are actually confident in their answers and find the

survey easy to complete, or the content of the survey may representimportant or highly accessible information (e.g., Fazio, 1990). Assuch, time spent on the survey may not predict ERS in and of itself,

but respondents who speed through the survey and are high in in-tolerance of ambiguity, simplistic thinking, or decisiveness might be

particularly likely to use an ERS. That is, quickly speeding throughthe survey may amplify the expected positive relations between in-

tolerance of ambiguity, simplistic thinking, and decisiveness withERS. By demonstrating this pattern for each of the three personality

measures, we will obtain more precise validation of the underlyingdispositional nature of ERS.

Methodological Issues in Assessing Extreme Response Style

Given the proposed relations between personality measures and ex-treme responding, it is important to ensure that the method of per-

sonality assessment used is both appropriate and valid. Because thepreviously specified measures intolerance of ambiguity, decisiveness,

and simplistic thinking all are traditionally assessed with self-reportLikert scales, these measures may be confounded by the extreme re-

sponses of those completing the items. Using self-reports to predictERS that are contaminated by the respondent’s own ERS will lead

only to equivocal findings.A possible method for addressing this problem is to use peer rat-

ings of personality variables in lieu of self-report predictors. By ask-

ing close friends of the respondents to assess the respondent’s level ofintolerance of ambiguity, simplistic thinking, and decisiveness, the

potential confounds of the respondent’s own extreme responses canbe avoided. Research has indicated that peer ratings of personality

constructs are both highly accurate and reliable alternatives for self-reports, with peer ratings even occasionally serving as better behav-


ioral predictors than self-report measures (Barbaranelli & Caprara,

2000; Barrick, Patton, & Haugland, 2000; Biesanz & West, 2000).Using peer reports of personality will produce responses to person-

ality predictors of ERS that are unconfounded by the original ERSof the respondent. Using peer reports also will eliminate the common

method bias associated with solely using self-report measures (Pods-akoff, MacKenzie, Lee, & Podsakoff, 2003).

A second issue involves developing an accurate measure of ERS.In this sense, the proposed study advances the past literature, which

has measured ERS in strikingly different ways. Researchers havemeasured extreme responses by counting only positive extreme re-sponses (Brengelmann, 1960a), by collapsing categories in longer

Likert scales (Borgatta & Glass, 1961), by subtracting midpoint usefrom endpoint use (see Hamilton, 1968), or even using standard de-

viation as a measure of extreme responding (see Hamilton, 1968).Given our proposition that ERS is a reflection of stable individual

differences, it makes sense that a measure of extreme respondingshould reflect this quality. This study will assess ERS in several

different ways, each of which should contribute to assessing the un-derlying core of response extremity.

In sum, it is proposed that ERS is a primarily stable dispositional

variable that is reflective of certain personality traits that differamong individuals, namely intolerance of ambiguity, decisiveness,

and simplistic thinking. It is also proposed that these personalityvariables will interact with time spent on survey to predict ERS, such

that those who complete the survey quickly and are high in thesetraits will be the most likely to use ERS. This study will also attempt

to determine the best way of measuring ERS in light of its relation-ship with these personality predictors. The current study thus marks

an improvement over past studies through the development of bettermeasures of the ERS construct and through its use of peer reports toremove the confounding bias of the respondents’ own extreme re-

sponding on the personality questionnaire predictors.

METHOD

Participants

The study sample consisted of 132 undergraduates at a large southernuniversity (primary participants), matched with 132 close friends (peer


participants). The undergraduates were taken from a larger sample of 522individuals (i.e., paired data including both self-reports and peer reportsexisted only for 132 of the original 522). The sample was 61% female,with ages ranging from 17 to 22 (M5 19.1, SD5 1.0). Regarding ethnic-ity, the sample contained 109 White participants, 2 African Americanparticipants, 13 Hispanic participants, 3 Asian participants, 2 participantswho marked ‘‘other,’’ and 3 participants who did not indicate any eth-nicity. The characteristics of this smaller sample were largely reflective ofthe characteristics of the full sample.

For the purposes of hypothesis testing, 16 participants who failed tofollow directions or respond to a majority of items were also removedfrom further analyses, leaving a final sample of 116 participants.2 Furtheranalyses involving time data also resulted in the removal of outliers (ap-proximately 7% of time data).

Procedure

Primary participants completed a battery of self-report questionnairemeasures in an online survey over approximately a 1-hour period, in ex-change for course credit. Participants were also encouraged to recruit peerrespondents who knew them well enough to rate them on all of the samequestionnaire measures (see below for more details). To provide an in-centive for the peer respondents, two $50 prizes were offered to two ran-domly selected peer respondents. Although the validity of the peers’school e-mail addresses was verified, no further eligibility criteria wereestablished to determine the closeness of peer acquaintance.

Materials for Primary Survey Participants

For the primary survey participants, the questionnaire battery consistedof three sections containing a series of separate measures presented inthree different orders for the purposes of varying the content location.Respondents in the sample were randomly assigned to one of these threeversions, with relatively equal proportions of respondents for each ver-sion of the questionnaire.

Demographics

Participants self-reported sex, ethnicity (converted to minority status),and SAT score (which served as a proxy for cognitive ability). These de-

2. These individuals were identified by examining responses to positively and

negatively keyed items to determine whether participants were responding ran-

domly.


mographic data were collected in order to control for any demographiceffects suggested by the past ERS literature.

Extreme Response Style

Three measures of ERS were collected toward the goal of converging onthe ERS construct. The first measure of ERS was a simple score consist-ing of the proportion of extreme response category (i.e., end points: 1 and6) use across 200 items in the questionnaire battery (excluding items usedin other measures of ERS and personality, as described below). Usingsuch a summative measure across all items in a battery is the most com-mon method of measuring ERS in the literature (Hamilton, 1968). Be-cause each item in the battery represents an opportunity for an extremeresponder to make an extreme response, using all the items in the scale,coded for response extremity, can provide a very large and reliable scaleto measure ERS behavior.

The second measure of ERS used was the so-called contentless mea-sure designed by Greenleaf (1992b) for the explicit purpose of measuringERS. Greenleaf’s measure accounts for a potential flaw of the first ERSmeasure by removing content as an influence on extreme responding,given that the first ERS measure might be tainted by multiple ‘‘pockets’’of items that are highly correlated. The Greenleaf scale incorporates 16items from an extensive survey administered to a large U.S. sample in1975 and 1987. Items were selected from these surveys based on their lowinter-item correlations and equivalent response category proportions,meaning that these items are essentially free of content influences and canbe considered representative of a respondent’s scale use on any six-cat-egory Likert scale. The proportion of extreme response scores on these 16items was also previously shown to be stable over a 12-year period, sup-porting that ERS is reflective of stable traits (Greenleaf, 1992b).

The third measure of ERS accounts for a potential flaw in Greenleaf’s(1992b) measure based on the fact that the items comprising this measurewere individually selected post hoc from various sections of a lengthyquestionnaire. Because presenting the 16 items of the Greenleaf scale inonly one location might result in differing inter-item correlations whencompared to Greenleaf’s original survey data, a third measure of extremeresponding was constructed using Greenleaf’s technique on the remaining273 items of this data set. By selecting 16 items from the current ques-tionnaire battery that are shown to have low inter-item correlations andsimilar response proportions, any potential problems with presenting theGreenleaf items in one particular location of the questionnaire batterycan be avoided.


Each of these methods of measuring ERS presumably captures re-sponse extremity in conceptually distinct ways.3 Whether these differingconceptual approaches to measuring ERS result in significantly differentmeasures of ERS is an empirical question, but it is expected that thesethree measures of ERS will converge on some general tendency for ex-treme responding.

Time

The online survey service provided a measure of time taken to completethe survey. This measure of time was calculated from data indicating theexact time each primary respondent both started and finished the onlinesurvey. As is the case with most time data, a set of outliers completed thesurvey over a much longer period of time, and consistent with other time-based studies (e.g., Ratcliff, 1993) times greater than three standard de-viations from the mean were dropped for the purposes of this analysis,resulting in the elimination of times for 7% of responders. It is likely thatthis small group of participants reflects respondents who interrupted theirsurvey-taking session and returned at a later time.

Personality Measures

Although the primary participants did complete self-ratings on Intoler-ance of Ambiguity, Simplistic Thinking, and Decisiveness, these scaleswere collected primarily to provide some assessment of the convergentvalidity of peer ratings on these traits. Because we analyzed the person-ality data using peer reports and not self-reports, we report details ofthese measures below in the Material for Peer Respondents section.

Materials for Peer Respondents

Each primary participant was asked to recruit a close friend to complete asimilar questionnaire. The questionnaire content for these peer ratingsconsisted of the same measures of personality (i.e., Intolerance of Am-biguity, Simplistic Thinking, and Decisiveness) described above. Peerparticipants, however, were instructed to complete the questionnaire as ifthey were the primary participant. Thus, peer ratings of each personalitycharacteristic were obtained for each primary participant. In the analysesreported below, we use the peer’s ratings of the primary participant’s

3. As discussed earlier, other measures of ERS exist. We decided against using

most of these as they contained obvious flaws. For example, measures of ERS that

count only positive responses confound trait level with response extremity.


personality to predict the participant’s manifestation of ERS (i.e., derivedfrom the primary participant questionnaire).

Intolerance of Ambiguity

A 25-item self-report questionnaire of intolerance of ambiguity was de-veloped that incorporated items from previous scales purporting to mea-sure intolerance of ambiguity (Budner, 1962; Norton, 1975; Rydell &Rosen, 1966; Webster & Kruglanski, 1994) as well as newly written items.All items were written such that higher scores indicate greater intoleranceof ambiguity (see Appendix). In addition, a previously used measure ofintolerance of ambiguity, the eight-item discomfort with ambiguity facetof the Need for Closure Scale (Webster & Kruglanski, 1994), was used forvalidation purposes. This measure was not used to test any hypotheses,however, because some item content overlapped with need for certainty.In addition to this issue, previous researchers have noted other constructcontamination issues with this measure (see Neuberg, Judice, & West,1997).

Simplistic Thinking

A new eight-item measure was developed for the purposes of capturingthe tendency to think in simple or dichotomized terms as opposed to thepreference for unambiguous or simple information (as assessed by thepreviously described Intolerance of Ambiguity measure) and items werewritten such that higher scores indicated greater tendency for simplicity(see Appendix). The scale employed new items as well as items from pre-vious measures of intolerance of ambiguity (Budner, 1962; Rydell &Rosen, 1966) and dogmatism (Rokeach, 1960).

Decisiveness

A new eight-item measure of decisiveness was developed with the inten-tion of focusing solely on the tendency to quickly ‘‘seize’’ on an answer, asopposed to including confidence or fear of making errors, given that thereis no theoretical reason to believe that these particular qualities might berelated to ERS. The newly developed eight-item measure incorporatessome items from the Personal Fear of Invalidity scale (Thompson et al.,1989) and the Need For Closure Scale (NFCS) decisiveness facet (Web-ster & Kruglanski, 1994), which were both also used for validation pur-poses but not hypothesis testing (given that these scales also measureconfidence and fear of making errors). The Decisiveness scale was written


such that higher scores indicate a greater tendency to seize on an answer(see Appendix).

RESULTS

See Table 1 for means, standard deviations, and correlations be-tween all predictors and measures of extreme response style. Relia-

bilities of all relevant measures are also reported on the diagonal ofthe matrix.

Scale Refinement

For each of the newly developed scales, exploratory factor and re-liability analyses were conducted to eliminate poor items and assess

convergent validity.


After removing 7 items, the resulting 25-item measure of ambiguityintolerance achieved a reliability of .93 and exhibited a strong single-factor solution in an exploratory factor analysis. This Intolerance of

Ambiguity measure was correlated .62 with the discomfort with am-biguity facet of the Need for Closure Scale (Webster & Kruglanski,

1994), indicating satisfactory construct validation with previousmeasures. Values on this scale ranged from 2.30 to 6.00 (M5 4.05,

SD5 0.67), and the correlation between self-reports and peer reportsof this measure was significant (r5 .52).

Simplistic Thinking

After removing seven items (most of which belonged to a separatefactor that was deemed less relevant), the resulting eight-item scale

was confirmed to be undimensional and achieved a reliability of .87.Values on this scale ranged from 1.25 to 6.00 (M5 3.52, SD5 0.96),

and the correlation between self-reports and peer reports of thismeasure was also significant (r5 .46).

Decisiveness

After removing two poor items, the resulting eight-item scale wasconfirmed to be unidimensional and achieved a reliability of .80.


Ta

ble

1M

ea

ns,

Sta

nd

ard

De

via

tio

ns,

an

dC

orr

ela

tio

ns

Be

twe

en

Pre

dic

tors

of

Ex

tre

me

Re

spo

nse

Sty

le(N

511

6)

MSD

12

34

56

78

910

11

12

13

14

1.Sex

—

2.Minority

Status

�.14

—

3.SATScore

1198.47

125.57�.22n�.05

—

4.Tim

e2783.26

1325.14�.03

.17

.05

—

5.PR

Ambiguity

4.05

0.67�.06�.10�.02�.12(.93)

6.PR

Sim

plistic

3.52

0.96�.13�.10

.10�.31n.65n(.87)

7.PR

Decisiveness

3.59

0.73�.17

.00

.08�.08

.15

.33n(.82)

8.ERSOverall

0.20

0.12

.03

.05

.19�.02

.29n.25n.25n

(.97)

9.ERSGreenleaf

0.26

0.19�.02

.10

.03�.04

.30n.12

.25n

.50n

(.67)

10.ERSAlternate

Greenleaf

0.20

0.15�.19

.09

.11�.04

.28n.21

.22n

.70n

.45n

(.71)

11.ERSUnit-w

eighted

0.22

0.13�.07

.10

.12�.04

.35n.22n.29n

.84n

.83n

.83n

(.76)

12.SR

Ambiguity

3.90

0.65�.06�.08�.11�.14

.52n.40n.18

.18

.14

.23n

.22n(.91)

13.SR

Sim

plistic

3.22

0.86�.22n�.09�.05�.17

.34n.46n.23n�.07�.04�.05�.06

.48n(.82)

14.SR

Decisiveness

3.36

0.78�.24n

.02

.07�.15

.10

.23n.55n

.17

.15

.29n

.24n.26n.33n(.80)

Note.Sex:0

5male,1

5female;Minority

status:0

5White,

15

non-W

hite;Ambiguity

5Intolerance

ofAmbiguity;Sim

plistic

5Sim

plistic

Thinking;PR

5PeerReport;SR

5Self-report;Coefficientalphasonthediagonal.

npo.05.

Evidence of convergent validity was exhibited by strong correlations

with the decisiveness facet of the NFCS (r5 .64) and the PersonalFear of Invalidity scale (Thompson et al., 1989; r5 � .62), which

measures indecisiveness. Values on this scale ranged from 1.56 to6.00 (M5 3.59, SD5 0.73), and the correlation between self-reports

and peer reports of this measure was also significant (r5 .55).

Extreme Response Style

As discussed above, we obtained the first measure of ERS by calcu-lating the proportion of endpoint use across the 200 items in the

questionnaire battery. Before recoding the items into endpoint usescores, the average inter-item correlation for the 200 items was .036,indicating a large amount of content variance. After recoding the

items for endpoint use (i.e., 1 for any endpoint category use, 0 foranything else), the internal consistency was exceedingly high

(a5 .97), indicating a substantial amount of systematic variance inendpoint use.

The second measure of ERS, the Greenleaf scale, had an averageinter-item correlation of .077 prior to recoding into the use of end-

point format, but some inter-item correlations ranged as high as .39,indicating some lack of content variance for a measure specificallydesigned to have very low inter-item correlations. The reliability of

the Greenleaf scale after coding for extreme responses was .71, againindicating a high degree of systematic variance due to endpoint use.

To compute the third measure of ERS, the alternate version of theGreenleaf scale, 16 items were selected from across the entire ques-

tionnaire battery (excluding the 16 original Greenleaf items anditems from the three personality predictor measures). These items

were deliberately chosen based on their low (o.1) correlations withone another and similar response proportions across categories (to

avoid highly skewed items). The average inter-item correlation of theAlternate Greenleaf scale was .004 before recoding into extreme re-sponses, with a reliability of .67 after recoding, also indicating a large

proportion of systematic variance due to endpoint use.The ERS measures were all substantially correlated with each

other, ranging from .45 to .70. Exploratory factor analyses indicatedthat the three measures of ERS all loaded highly onto a single factor

(loadings 4.80). As such, a single unit-weighted ERS score was cal-culated based on the mean of the ERS scores of the overall ques-


tionnaire battery, the Greenleaf scale, and the Alternate Greenleaf

scale. This unit-weighted ERS score was used as the primary mea-sure of ERS for subsequent hypotheses, with a scale reliability of .76.

ERS scores ranged from 0 to .78, with a mean of .22, and a standarddeviation of .13. This indicates that approximately 22% of responses

were extreme.4

Stability of ERS

The stability of ERS across the survey battery was assessed by ex-amining whether the order in which the questionnaire scales were

presented affected the likelihood of extreme responding. Participantscompleted the survey battery in one of three different section orders,

and ERS scales were formulated to appear either at the beginning,middle, or end of the questionnaire battery based on the particular

version of the questionnaire battery given to the sample. Results in-dicated that ERS scores did not depend upon the location of theitems within the survey (b5 .02, p4.05).

Regressions

Initial support for our hypotheses can be seen in Table 1. Intoleranceof ambiguity, simplistic thinking, and decisiveness were all positivelyand significantly related to ERS.

Given past research demonstrating relations between demo-graphic variables and ERS and potential multicollinearity issues

with our predictors, we examined the predictors separately control-ling each time for sex, minority status, and self-reported SAT score.

None of these demographic control variables were found to predictERS (see Table 2).

Intolerance of ambiguity, controlling for sex, minority status, andself-reported SAT score, accounted for 15.5% of the variance in

ERS. Simplistic thinking, controlling for the same demographicvariables, accounted for 7.5% of the variance in ERS. Finally, de-cisiveness, controlling for demographic variables, accounted for

10.3% of the variance in ERS.

4. We say ‘‘approximately,’’ as this unit-weighted composite does not treat each

item equally (i.e., there are only 16 items in each version of the Greenleaf scale,

whereas the overall scale has 200). The actual overall proportion of endpoint use

was 20%, with a SD of 13%.


Intolerance of ambiguity, simplistic thinking, and decisiveness ac-count for 24.1% of the variance in ERS simultaneously, when con-

trolling for demographic variables (b5 .43, po.05, for ambiguity,b5 .32, po.05, for decisiveness, b5 � .10, p4.05, for simplisticthinking). These results show that the effect of simplistic thinking on

Table 2Regressions and Interactions (N 5116)

Personality variable b DR2 R2


Step 1.

Gender .03

Minority status � .03

SAT score .13 .02 .02

Step 2.

Time .04

Intolerance of Ambiguity .39 .15n .17

Step 3.

Time � Intolerance of Ambiguity � .25 .05n .22

Simplistic Thinking

Step 1.

Gender .03


SAT score .13 .02 .02

Step 2.

Time .09

Simplistic Thinking .29 .07n .09

Step 3.

Time � Simplistic Thinking � .54 .13n .22

Decisiveness

Step 1.

Gender .03


SAT score .13 .02 .02

Step 2.

Time � .01

Decisiveness .29 .08n .10

Step 3.

Time � Decisiveness � .14 .02 .12

npo05.


ERS is eliminated when the other personality variables are also in-

cluded, a result that is unsurprising given the high correlation be-tween simplistic thinking and intolerance of ambiguity (r5 .65,

po.05).

Interactions

Moderated multiple regression was used to test interactions betweenthe personality measures and time spent on the survey. In Step 1, thecontrol variables sex, minority status, and SAT score were entered.

In Step 2, main effects for time and personality were entered. Finally,interactions between time and personality were entered in Step 3. All

main effects were centered prior to forming interaction terms. Theseresults are displayed in Table 2.

Results showed significant interactions for both intolerance ofambiguity (b5 � .25, po.05; Figure 1) and for simplistic thinking

(b5 � .54, po.05; Figure 2). There was no significant interaction fordecisiveness (b5 � .14, p5 .21).

Figures 1 and 2 support two of the three interaction hypotheses by

showing that likelihood of extreme responding increases when timespent on survey is low and scores on the intolerance of ambiguity

and simplistic thinking measures are high and remains relativelystable when time spent on the survey is high. Both interactions dem-

onstrate that those who are most likely to engage in extreme re-sponding are those who rush through the survey and are also high in

the specified personality measure.

0

0.1

0.2

0.3

0.4

0.5

Low Intolerance ofAmbiguity

High Intolerance ofAmbiguity

ER

S

Low Time Spent onSurveyHigh Time Spent onSurvey

Figure 1Interactive effect of time spent on survey on the relation between

intolerance of ambiguity and ERS.


DISCUSSION

The goal of this study was to understand the nature of relations be-tween personality and extreme response style. All three hypothesized

personality constructs were found to be positively related to ERS,and these overall effects were stronger for people who completed the

survey more quickly. These effects also occurred over and above anydemographic predictors of ERS.

In terms of demographic predictors of ERS, past studies had in-dicated mixed support for the possibility of sex and cognitive ability

predicting ERS and some support for ethnicity predicting ERS. Wefound no differences between sexes on ERS. Cognitive ability wasnot significantly related to ERS; however, self-reported SAT scores

may not have been a valid enough measure of cognitive ability. Inaddition, our college sample likely yielded some range restriction.

The lack of findings for ethnicity as a predictor of ERS may reflect atrue lack of ethnic differences in ERS, but considering the limited

sample of non-White participants, this conclusion is only tentative.Past studies also suggested a positive relationship between con-

structs similar to intolerance of ambiguity and ERS, and this studyfound that a measure with item content specifically devoted to in-tolerance of ambiguity (and not the host of other constructs confla-

ted in past measures of intolerance of ambiguity) accounted for asignificant amount of variance in extreme responding. This relation-

ship held after accounting for other demographic and personalitypredictors of ERS. As such, these results indicate that a measure

0

0.1

0.2

0.3

0.4

0.5

Low SimplisticThinking

ER

SLow Time Spent onSurvey

High Time Spent onSurvey

High SimplisticThinking

Figure 2Interactive effect of time spent on survey on the relation between

simplistic thinking and ERS.


whose item content is designed to tap only intolerance of ambiguity

is related to extreme responding.An additional contribution of this study was the uncovering of a

significant relationship between decisiveness and ERS. Specifically, atendency to quickly seize on the strongest possible answer was shown

to account for a significant amount of variance in extreme respond-ing, above and beyond control variables, and also when included

with other personality variables.Finally, a novel positive relationship between a tendency for sim-

plistic thinking and ERS was uncovered. Because simplistic thinkingheld a strong relation to intolerance of ambiguity, this effect waseliminated when all three personality variables were included in the

regression model predicting ERS. Taken together, however, theseresults indicate that at least three personality variables that can be

grouped together as stable epistemic constructs all predict variancein ERS, supporting the stable and dispositional nature of ERS.

In terms of contextual moderators of ERS, two significant inter-actions were discovered with response time. Specifically, participants

who finished the survey quickly and were rated as high in ambiguityintolerance (as well as simplistic thinking) were far more likely toexhibit an extreme response style than those who were rated low in

these personality variables by peers. This pattern also held for de-cisiveness: Although the interaction was not significant, the direction

of the relationship was such that respondents who rushed throughthe survey and were rated as highly decisive were the most likely to

exhibit ERS.In short, simply rushing through a questionnaire is not sufficient

to lead to ERS; one must also be highly intolerant of ambiguity,decisive, or inclined toward simplistic thinking. These results make

sense, as it is possible for those who rush through a survey to chooseto respond randomly, or respond with a central tendency. If some-one is high in the aforementioned personality traits, however, then

they will likely employ an extreme response style when completingthe survey quickly.

Based on the success of using a combination of three differentmeasures of ERS, we can offer several suggestions for researchers

who wish to assess ERS in their own surveys. First, given that eachmeasure of ERS is likely to contain at least one drawback, if there is

sufficient space in a survey, then researchers should use multiplemeasures to triangulate on participants’ latent ERS. If space


considerations require a more abbreviated approach to ERS assess-

ment, then we would recommend using the method described byGreenleaf (1992b) and used for our alternative Greenleaf scale. Spe-

cifically, one could select a set of items from the entire survey that arerelatively uncorrelated and have similar response proportions. Then

score these items according to the use of endpoints and use the meanas the individual’s ERS score. In Greenleaf’s study and again in our

study, a relatively small selection of items created a fairly reliableERS scale (i.e., 16 items in both cases) that contained minimal over-

lap in content. In contrast to this approach and despite the high levelof internal consistency for our overall ERS scale, we would not rec-ommend using the combined proportion of extreme responses across

all items as the sole assessment of ERS. In particular, it seems likelythat a good deal of content overlap will exist in any survey, partic-

ularly given that researchers often have the goal of including scalesthat are highly related to one another. In such cases ERS is likely to

be confounded with trait level on one or more measured constructs.A final consideration in assessing ERS is whether material placed

at the end of a survey is more likely to exhibit ERS than earlierportions of the survey (Hui & Triandis, 1985). The current studyfound no support for this notion, as the order in which ERS mea-

sures were presented had no significant effect on likelihood of ex-treme responding.

The accumulated evidence in this study indicates that althoughERS may be influenced by situational factors, a substantial amount

of the variance in extreme responding is due to stable personalitytraits. This conclusion is buttressed by support from moderate rela-

tions between ERS and three personality variables, evidence that therelationship between these personality variables and ERS grows

stronger for people who spend less time completing the survey andthe lack of evidence for fatigue or speed as an explanation for ERS.

Despite the consistency of these results, there are, of course, sev-

eral limitations to be noted. The use of a college sample begs thequestion of whether the effects are the same for those who are less

educated or experienced with completing surveys. Also, our use ofelectronic survey methodology may not generalize to surveys ad-

ministered using paper and pencil.Although our new measures of the three epistemic personality con-

structs exhibited generally high levels of reliability and construct validity,the lack of prior validation leaves some room for caution in the inter-


pretation of our results. Hopefully, extension of these newly developed

personality measures in other contexts will prove useful and will helpsolidify the results obtained here. A final limitation that we should men-

tion concerns our attempts to rule out the effects of demographic vari-ables. Our conclusions concerning the role of ethnicity and intelligence in

particular will remain tenuous until replication can occur with a morediverse sample and a more valid measure of cognitive ability.

Despite these limitations, the results of this study are strengthenedby several features of our research design, including the use of peer

ratings of personality in lieu of self-reports, multiple measures forboth predictor and criteria constructs, and the examination of sev-eral control variables. The implications of these results have bearing

in any situation that requires the rating of a stimulus or individual,as it appears that personality factors play a key role in whether or

not an extreme response is chosen over a nonextreme response, ir-respective of the actual content of the stimulus. In situations where

cautious judgments are preferable or where the consequences of aninaccurate extreme judgment may be severe (such as in clinical or

medical decisions), it may be advisable to determine whether therater possesses personality characteristics associated with an ERS.

Although our study does not offer direct implications for elimi-

nating potential biases associated with ERS, we again offer somepossibilities for future research to examine. Statistical control based

on the direct assessment of ERS is one such possibility. For example,if ERS is thought to influence criterion scores, then researchers

might first enter ERS in the regression model, relying afterward onthe residual criterion scores. Unfortunately, this method only ad-

dresses a relatively simplistic linear effect of ERS. A second methodwould be to derive new norms for extreme and nonextreme respond-

ers in popular personality inventories such as tests based on the Five-Factor Model of personality.

Understanding response styles such as ERS is not only important

from a psychometric or practical standpoint, but also interesting as atheoretical investigation of a dispositional style that may influence

behavior in substantive areas of work (Guion, 1998). Individualswho are likely to make extreme responses when using Likert scales in

questionnaires could also possibly be driven to make extreme re-sponses in general social situations (e.g., political attitudes, decision

making, group interactions). Casting ERS as a behavioral manifes-tation of intolerance of ambiguity, decisiveness, and simplistic think-


ing also serves as a form of validation for these epistemic personality

constructs, based on the hypothesized links between dispositionaltraits and behavioral actions. In this sense, the construct of ERS can

be considered both as an indicator of certain personality traits aswell as a possible measure of bias. Future studies might therefore

benefit by examining whether other personality variables, such as theFive-Factor Model of personality, will account for additional unique

variance in extreme responding. In this way, studying ERS from apersonality perspective should hopefully contribute to this underre-

searched yet widely applicable topic.

REFERENCES

Austin, E. J., Deary, I. J., & Egan, V. (2006). Individual differences in response

scale use: Mixed Rasch modelling of responses to NEO-FFI items. Personality

and Individual Differences, 40, 1235–1245.

Bachman, J., & O’Malley, P. (1984). Yea-saying, nay-saying, and going to ex-

tremes: Black white differences in response styles. Public Opinion Quarterly, 48,

491–509.

Barbaranelli, C., & Caprara, G. (2000). Measuring the big five in self-report and

other ratings: A multitrait-multimethod study. European Journal of Psycho-

logical Assessment, 16, 31–43.

Barrick, M., Patton, G., & Haugland, S. (2000). Accuracy of interviewer judgment

of job applicant personality traits. Personnel Psychology, 53, 925–952.

Biesanz, J., & West, S. (2000). Personality coherence: Moderating self-other pro-

file agreement and profile consensus. Journal of Personality and Social Psy-

chology, 79, 425–437.

Berg, I., & Collier, J. (1953). Personality and group differences in extreme re-

sponse sets. Educational and Psychological Measurement, 13, 164–169.

Borgatta, E., & Glass, D. (1961). Personality concomitants of extreme response

set (ERS). Journal of Social Psychology, 55, 213–221.

Brengelmann, J. (1960a). Extreme response set, drive level and abnormality in

questionnaire rigidity. Journal of Mental Science, 106, 171–186.

Brengelmann, J. (1960b). A note on questionnaire rigidity and extreme response

set. Journal of Mental Science, 106, 187–192.

Brim, O., & Hoff, D. (1957). Individual and situational differences in desire for

certainty. Journal of Abnormal and Social Psychology, 54, 225–229.

Budner, S. (1962). Intolerance of ambiguity as a personality variable. Journal of

Personality, 30, 29–59.

Crandall, J. (1973). Sex differences in extreme response style: Differences in fre-

quency of use of extreme positive and negative ratings. Journal of Social Psy-

chology, 89, 281–293.

Das, J., & Dutta, T. (1969). Some correlates of extreme response set. Acta Psy-

chologica, 29, 85–92.


Eid, M., & Rauber, M. (2000). Detecting measurement invariance in organiza-

tional surveys. European Journal of Psychological Assessment, 16, 20–30.

Fazio, R. (1990). Multiple processes by which attitudes guide behavior: The MOD

model as an integrative framework. In M. Zanna (Ed.), Advances in experi-

mental social psychology (Vol. 23, pp. 75–109). New York: Academic Press.

Gerardo, M., Gamba, R., & Marin, B. (1992). Extreme response style and ac-

quiescence among Hispanics: The role of acculturation and education. Journal

of Cross Cultural Psychology, 23, 498–509.

Greenleaf, E. (1992a). Improving rating scale measures by detecting and correct-

ing bias components in some response styles. Journal of Marketing Research,

29, 176–188.

Greenleaf, E. (1992b). Measuring extreme response style. Public Opinion Quar-

terly, 56, 328–351.

Guion, R. (1998). Assessment, measurement and prediction for personnel decisions.

Mahwah, NJ: Erlbaum.

Hamilton, D. (1968). Personality attributes associated with extreme response

style. Psychological Bulletin, 69, 192–203.

Hui, C., & Triandis, H. (1985). The instability of response sets. Public Opinion

Quarterly, 49, 253–260.

Jain, U., & Agrawal, L. (1977). Generality of extreme response style. Journal of

Psychological Researches, 21, 67–72.

Johnson, T. (2003). On the use of heterogeneous thresholds ordinal regression

models to account for individual differences in response style. Psychometrika,

68, 563–583.

Johnson, T., Kulesa, P., Cho, Y., & Shavitt, S. (2005). The relation between

culture and response styles: Evidence from 19 countries. Journal of Cross-

Cultural Psychology, 36, 264–277.

Kerrick, J. (1954). The effects of intelligence and manifest anxiety on attitude change

through communications. Unpublished doctoral dissertation, University of Illinois.

Kruglanski, A. (1989). Lay epistemics and human knowledge: Cognitive and mo-

tivational bases. New York: Plenum.

Lewis, N., & Taylor, J. (1955). Anxiety and extreme response preferences. Edu-

cational & Psychological Measurement, 15, 111–116.

Light, C., Zax, M., & Gardiner, D. (1965). Relationship of age, sex, and intel-

ligence level to extreme response style. Journal of Personality & Social Psy-

chology, 2, 907–909.

Merrens, M. (1970). Generality and stability of extreme response style. Psycho-

logical Reports, 27, 802.

Mischel, W., & Shoda, Y. (1995). A cognitive-affective system theory of person-

ality: Reconceptualizing situations, dispositions, dynamics, and invariance in

personality structure. Psychological Review, 102, 246–268.

Molto, J., Segarra, P., & Avila, C. (1993). Impulsivity and total response speed

to a personality questionnaire. Personality and Individual Differences, 15,

97–98.

Neuberg, S., Judice, T., & West, S. (1997). What the Need for Closure Scale

measures and what it does not: Toward differentiating among related epistemic

motives. Journal of Personality and Social Psychology, 72, 1396–1412.


Norton, R. (1975). Measurement of ambiguity tolerance. Journal of Personality

Assessment, 39, 607–619.

Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003).

Common method biases in behavioral research: A critical review of the

literature and recommended remedies. Journal of Applied Psychology, 88,

879–903.

Ratcliff, R. (1993). Methods for dealing with reaction time outliers. Psychological

Bulletin, 114, 510–532.

Roets, A., & Van Hiel, A. (2007). Separating ability from need: Clarifying the

dimensional structure of the Need for Closure Scale. Personality and Social

Psychology Bulletin, 33, 266–280.

Rokeach, M. (1960). The open and closed mind. New York: Basic Books.

Rydell, S., & Rosen, E. (1966). Measurement and some correlates of need-cog-

nition. Psychological Reports, 19, 139–165.

Soueif, M. (1958). Extreme response sets as a measure of intolerance of ambiguity.

British Journal of Psychology, 49, 329–334.

Thompson, M., Naccarato, M., & Parker, K. (1989, June). Assessing cognitive

need: The development of the Personal Need for Structure and Personal Fear of

Invalidity Scales. Paper presented at the annual meeting of the Canadian Psy-

chological Association, Halifax, Nova Scotia, Canada.

Webster, D., & Kruglanski, A. (1994). Individual differences in need for cognitive

closure. Journal of Personality and Social Psychology, 67, 1049–1062.

Wilkinson, E. (1970). Relationship between measures of intellectual functioning

and extreme response style. Journal of Social Psychology, 81, 271–272.

Zuckerman, M., & Norton, J. (1961). Response set and content factors in the

California F Scale and the Parental Attitude Research Instrument. Journal of

Social Psychology, 53, 199–210.

APPENDIX

Items marked with an asterisk also appear in previously validated

measures. NFCS5Need for Closure scale (Webster & Kruglanski,1994), PFIS5Personal Fear of Invalidity scale (Thompson et al.,

1989), RDS5Rokeach Dogmatism scale (Rokeach, 1960),NAIS5Norton Ambiguity Intolerance Scale (Norton, 1975),

BS5Budner scale (Budner, 1962), RRAS5Rydell and Rosen Am-biguity scale (Rydell & Rosen, 1966).


1 (strongly disagree), 2 (disagree), 3 (slightly disagree), 4 (slightlyagree), 5 (agree), 6 (strongly agree)


1. Thinking about complex brainteasers is a waste of time.

2. I dislike questions which could be answered correctly in manydifferent ways.n (NFCS)

3. I find it annoying when someone says something that is overlycomplex.

4. I have no use for people who have ‘‘nuanced’’ opinions.5. I don’t like things that have more than one interpretation.

6. I like things that are plain and easy to understand.7. The simple life is the good life.

8. I prefer simple problems over complex problems.9. I like it when issues are black and white rather than ‘‘shades of

grey.’’

10. I like it when there are no ifs, ands, or buts about it.11. Poems with contradictions are annoying.

12. I dislike things that have both positive and negative qualitiesabout them.

13. Groups which have too much difference of opinion amongtheir own members don’t appeal to me.

14. I like stories that have consistent characters.15. Vague and impressionistic pictures really have little appeal for

me.n (RRAS)

16. I don’t like to work on a problem unless there’s a possibilityof coming out with a clear cut answer.n (RRAS)

17. It’s annoying to listen to someone who cannot seem to makeup his or her mind.n (NFCS)

18. I hate it when you can’t solve a problem right away.19. I hate it when a TV series ends its season with a cliffhanger.

20. A group meeting functions best with a definite agenda.n (NAIS)21. The best part of working a jigsaw puzzle is putting in that last

piece.n (RRAS)22. I like movies with clear cut endings.23. A good job is one where what is to be done is always clear.n (BS)

24. I usually get a strong sense of relief when I finally get the answer.25. I’d rather find an answer, any answer, compared to uncertainty.

Decisiveness

1. I may struggle with a few decisions, but not very often.n (PFIS)2. I never put off making important decisions.n (PFIS)


3. The possibility that I might be wrong rarely stops me from

making a decision.4. I tend to make my decisions immediately.

5. When faced with a problem I usually see the one best solutionvery quickly.n (NFCS)

6. I usually make important decisions quickly.7. I would describe myself as decisive.

8. He who hesitates is lost.

Simplistic Thinking

1. Of all the different philosophies which exist in this world thereis probably only one which is correct.n (RDS)

2. There are two kinds of people in the world: those who are forthe truth and those who are against the truthn (RDS)

3. There are only a few types of people in this world.4. There’s a right way and a wrong way to do almost everything.n

(RRAS)5. Most questions can be answered with a simple yes or no.

6. An expert who doesn’t come up with a definitive answer prob-ably doesn’t know too much.n (BS)

7. Practically every problem has a solution.n (RRAS)

8. I tend to see most issues in black and white terms.


Personality and Trust Fosters Service Quality

Documents

Transcript of Personality and Trust Fosters Service Quality